Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Xander44-4/traffic_reducer/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you from a freshly cloned repository to a live, AI-controlled traffic signal dashboard running in your browser. By the end you will have the Flask server running, a video source connected, and YOLOv8 actively counting vehicles and selecting green-light phases in real time.
Python 3.10 or 3.11 is recommended. Some packages in requirements.txt — particularly ultralytics and scikit-learn — may have compatibility issues on Python 3.12 or later. Run python --version before you begin to confirm your environment.
1

Clone the repository and navigate to the project

Clone the Traffic Reducer repository and change into the project root. All subsequent commands assume you are running from this directory.
git clone https://github.com/Xander44-4/traffic_reducer.git
cd traffic_reducer
The project root contains the YOLO weight files (yolov8m.pt, yolov8s.pt, yolov8n.pt) and the traffic_app/ directory with app.py and video_processor.py.
2

Install all Python dependencies

Install every required package in one command using the provided requirements file:
pip install -r requirements.txt
This installs the following packages:
PackageWhat it provides
flaskWeb server and REST API
opencv-pythonFrame capture and image processing
ultralyticsYOLOv8 inference (also installs imageio-ffmpeg)
yt-dlpYouTube live stream URL extraction
pandasData handling utilities
numpyNumerical operations and zone detection
joblibAlternative .pkl model loader
scikit-learnPre-trained signal-phase model compatibility
Installation typically takes one to three minutes depending on your internet speed. ultralytics is the largest dependency as it pulls in PyTorch.
3

Start the Traffic Reducer server

Launch the Flask application from the project root:
py traffic_app/app.py
On macOS or Linux, use:
python3 traffic_app/app.py
You will see the following startup output in your terminal:
[App] Modelo YOLO: yolov8m.pt
[App] Video local: .../traffic_dron_view.mp4 (OK)
[App] Modo por defecto: idle
Iniciando Traffic Reducer Server...
http://127.0.0.1:5000
The server starts in idle mode by default — the TrafficCamera background thread is running but no video source is active yet. The YOLO model is loaded and ready; video processing begins as soon as you choose a source in the next steps.
You may see TLS or Socket error messages in the console after the server starts. These are harmless. TrafficCamera includes automatic reconnection logic that handles transient stream failures, so the system will recover without any action on your part.
4

Open the dashboard in your browser

With the server running, open your web browser and navigate to:
http://127.0.0.1:5000
You will see the Traffic Reducer dashboard. The video panel shows a placeholder frame with the message “Selecciona una fuente para comenzar” because the system is in idle mode. The zone count panels (Norte, Sur, Este, Oeste) display zeroes until a video source is connected.
5

Choose a video source and activate AI control

  1. Click the source modal button in the dashboard toolbar. A dialog appears offering two options:
    • YouTube live stream — connects to the configured live traffic camera via yt-dlp and streams it through FFmpeg. An internet connection is required.
    • Local video — plays traffic_app/static/traffic_dron_view.mp4 at 0.5× speed from disk, with no internet dependency.
  2. Select your preferred source and confirm. The POST /set_source API endpoint switches TrafficCamera to the chosen mode. The console will print the connection status — for YouTube mode you will see: 
    [TrafficCamera] URL obtenida con cliente android
[TrafficCamera] Stream YouTube abierto (ffmpeg pipe) — esperando primer frame...
    When frames start arriving, the console will print [TrafficCamera] Stream YouTube abierto (ffmpeg pipe) — esperando primer frame... followed by frame processing logs. This confirms that YOLOv8 is actively processing frames and zone counts are being updated.
  3. Click “Activar Control” (or the LIVE SIMULATION MODE toggle) to enable AI decision mode. The system now calls POST /predict with live_mode: true on each cycle, applies np.argmax across the four zone counts, and sets the winning direction as the active green-light phase. The dashboard signal display updates in real time to reflect the AI decision.

​
What to expect

Once a source is active and AI control is enabled, the dashboard updates every second with:
  • Annotated video — bounding boxes colour-coded by zone (amber for Norte, cyan for Sur, green for Este, purple for Oeste), plus yellow boxes for pedestrians and red boxes for detected emergency vehicles
  • Zone vehicle counts — live counts for Norte, Sur, Este, and Oeste lanes
  • Active phase — the zone currently holding the green light, as decided by the majority-rule algorithm
  • Priority flags — a pedestrian or emergency indicator when an override is in effect
From here you can explore the REST API to integrate Traffic Reducer with external systems, or read the Architecture guide to understand how the detection pipeline and threading model work under the hood.

Build docs developers (and LLMs) love

Get started for freeTalk to us